Professional Documents
Culture Documents
DOI: 10.1039/c5fd90046h
Roger Newman responded: I never said that the passive lm is not involved in
pit initiation. I understand (of course) that longer passivation gives a longer
induction time for pitting. What I say is that the effects of parameters like alloy
composition, environment composition, potential, and temperature, are not
easily accommodated (at least not predictively) within a passive lm breakdown
model, but fall out naturally from a modied Galvele type of model that uses
pitting potential data (or, if one has the time, lower extremes of pitting potential
distributions). Now I don't know whether or not the nest details of lm break-
down, detectable at the pA level or lower in electrochemical experiments (not nA
to mA – those are already pits), and/or on very pure, at alloy surfaces, follow the
same rules that we nd using pitting potentials on industrial or semi-industrial
alloys. Those measurements have not been done. Actually I don't think stainless
steel is necessarily the best model system for such studies. Under certain
conditions, as shown by Bardwell many years ago, iron shows blizzards of pits
that are clearly not impurity-particle-related; probably aluminum too. In stainless
steel we really don't know whether pit initiation ever occurs without a micro-
crevice and/or an impurity particle.
From the viewpoint of practical utility, the Galvele type of approach clearly has
the advantage. The Critical Pitting Temperature (CPT) is a propagation-related
transition below which metastable pits never become stable at any potential. My
group has published extensively on that. One can make a foolproof, if expensive,
† Electronic supplementary information (ESI) available: SVET scans movie. See DOI: 10.1039/c5fd90046h
John Scully asked: Your paper discusses several important dissolution nuances
such as the way Fe and Ni dissolve during corrosion and how that is affected by
sulfur. Has our understanding advanced regarding how a nickel or iron atom
oxidizes given different rst and second nearest neighbor arrangements such as
in multi-element solid solutions?
Roger Newman replied: We are dealing here with what Pickering would have
called region 3 dealloying – that is, Ni enriches during simultaneous dissolution,
without (as far as we know) porosity formation. Some insights have been gained on
Published on 01 July 2015. Downloaded on 01/07/2015 16:00:38.
Pickering's region 1 dealloying (low potential, only one element dissolving – not
really accessible for FeNi) through work of Sieradzki and (for an initially ordered
alloy) Renner. Also, Erlebacher's MESOSIM simulation can change the “bond
energies” and examine the resulting kinetics and morphology.1–4 But simultaneous
dissolution is harder to study. Clearly Ni dissolution limits, in a way, the dissolution
of the whole alloy, and if the Ni dissolution is activated by S adsorption, then the
alloy will dissolve faster (of course Fe dissolution is also activated). Regarding the
effect of Ni content on this surface enrichment and its consequences, I don't believe
a systematic study has been done for the FeNi system, and surface analysis of dis-
solved or dissolving samples is a huge challenge. Obviously one doesn't get exactly
the same polarization curve for 5% Ni as for 50% Ni, and when I tried this (on a
small scale) I remember having problems with martensite at certain Ni content.
Experimentally, it would be nice if there was a binary alloy of more noble metals that
could be subjected to either simultaneous or selective dissolution depending on the
potential, with no interference from hydrogen evolution, but the usual Au alloys in
simple acids are not suitable for that, and many other systems have bad phase
diagrams. Possibly one could use a complexing solution like cyanide.
Many years ago we thought we had shown that porosity was formed during
simultaneous dissolution of FeNi in HCl solution, and we made a link with Cl-
SCC of stainless steels. I still believe that, although the papers were not very
convincing.5
Gerald Frankel said: Others have shown that the semiconducting properties on
the passive lm on Alloy 800 are a function of the thiosulfate concentration.1 This
behavior correlates with the pitting behavior. In a sense, your measurements are
also correlations. How can you rule out the impact of the passive lm?
Philippe Marcus asked: Do you have any direct evidence that pitting can be
initiated with sulfate/thiosulfate without chlorides (or is scratching necessary to
Published on 01 July 2015. Downloaded on 01/07/2015 16:00:38.
depassivate the surface)? In work we did on FeCr alloys1 we showed that when
both chlorides and thiosulfates are present, passivity breakdown is caused by
chlorides and then repassivation is precluded by adsorption of sulphur.
1 C. Duret-Thual, D. Costa, W. P. Yang and P. Marcus, The role of thiosulfates in the pitting
corrosion of Fe-17Cr alloys in neutral chloride solution: Electrochemical and XPS study,
Corros. Sci., 1997, 39, 913–933.
Roger Newman replied: It does initiate eventually if you just leave the sample
in the susceptible range of potential – that's one reason why we apply an initial
passivation in pure sulfate solution (the other reason is that we want a nice low
baseline current when we scratch). It does take longer to initiate than when we
have thiosulfate and chloride, which can be studied using ordinary voltamme-
try.1To study the sulfate–thiosulfate pitting using voltammetry requires an
extremely low scan rate.
I would emphasize (and this is very important) that sulfate–thiosulfate pitting
occurs at potentials that are far lower than where ordinary chloride pitting occurs.
Of course we don't know how much of the above is affected by particulate in
the alloy, or surface roughness. It may be useful to compare the chloride–thio-
sulfate and sulfate–thiosulfate systems using very clean alloys, to see if the
difference in initiation rate between chloride and sulfate increases. Choice of
surface nish(es) would be important in such experiments. It is possible that the
sulfate–thiosulfate pitting (or for that matter chloride–thiosulfate pitting) may
vanish on surfaces that are extremely smooth and particle-free (see my response
to Trevor Rayment for a comment on sputtered thin lms and the absence of
pitting above a certain Cr content).
Dirk Engelberg asked: I've noticed that all tested microstructures were in the
solution annealed condition, which is different to how these alloys are used in
plant (e.g. thermally treated – TT). Would you expect an effect of microstructure
on pit nucleation behaviour and frequency? For example, how would the presence
of carbides change the lower/upper bounds of pitting in Table 3/Figure 4?
Trevor Rayment commented: This is a question for Roger Newman and Phil-
ippe Marcus.
There has been a long standing debate in the corrosion community concerning
the mechanism by which molybdenum confers protection against pitting. A
simple, if not simplistic, summary of the two schools of thought is a follows. The
rst school (Philippe Marcus) proposes that the role of Mo is to make the passive
lm more resistant to pitting, whereas the second school (Roger Newman)
proposes that the role of Mo is upon dissolution kinetics aer initiation of pitting
has taken place.
The debate over the mechanism(s) of pitting protection of molybdenum is not
resolved but there has been no discussion of this important topic within this
meeting yet. A great deal of experimental data has been published but it does not
seem to have convinced all parties to the discussion. A Faraday meeting presents
an ideal forum for such a debate.
Therefore I would like to ask Roger and Philippe to propose experimental
evidence (new experiments) that others could provide which would persuade
them to change their mind on the mechanism for molybdenum protection in
corrosion.
Roger Newman answered: I believe many of the critical experiments have been
done already. The problem, from where I stand, is that a long list of phenomena
in pitting are nicely accounted for (and in several cases quantitatively) by a
modied Galvele type of model.1 Similar success was achieved in studies that
addressed the effects of bromide, thiosulfate, temperature, etc. The only thing
missing from this long list of successes is that (almost always) our denition of
pitting was stable pitting. If one's denition is “rst onset of detectable current
transients due to passive lm breakdown,” then indeed there is some evidence
that thiosulfate takes advantage of spontaneous breakdown events, detectable at
the nA level in pitting experiments. But these (perhaps chloride-induced) events
are still pits – quite large actually. Burstein and his students looked down to the
pA level and found what they considered to be a different type of event, due
perhaps to salt forming under the passive lm and bursting it open.2 I don't recall
whether the frequency and potential dependence of such events was compared in
steels with and without Mo. Such data (although the charges under the transients
were still substantial in terms of number of atoms dissolved) may be valuable in
determining whether or not Mo affects the nucleation frequency of relevant
issue here is that we do not know whether the pitting that we studied
(in commercial or home-made alloys) was related to second-phase particles.
1 N. J. Laycock and R. C. Newman, Localized dissolution kinetics, salt lms and pitting
potentials, Corros. Sci., 1997, 39, 1771–1790.
2 S. P. Mattin and G. T. Burstein, Philos. Mag. Lett., 1997, 76, 341.
3 M. P. Ryan, N. J. Laycock, R. C. Newman and H. S. Isaacs, The pitting behavior of Fe–Cr
thin lm alloys in hydrochloric acid, J. Electrochem. Soc., 1998, 145, 1566–1571.
John Scully enquired: Operating on the assumption that the pitting potential
reects ease of initiation and stabilization with perhaps more weight on stabili-
zation, are there examples where the oxide plays a big role in initiation and
Published on 01 July 2015. Downloaded on 01/07/2015 16:00:38.
thereby affects the pitting potential? In relatively defect free materials, could
emphasis shi towards the oxide? One observation has been in the case of
amorphous glasses and high purity or relatively defect free stainless steels that
there are very few initiation events detected at the instrument limit. These alloys
enjoy very high pitting potentials and it seems that the rst “detected pit”
immediately stabilizes pitting. This stands in contrast to garden variety stainless
steels and aluminum alloys with many inclusions and intermetallics where there
are many initiation events. You have mentioned that if the defect is on the order
of the oxide thickness then stabilization would require extraordinary anodic
currents and I agree. But, what was the atomistic event that led to rupture of the
oxide in the rst place and is oxide engineering then a fruitful avenue for research
in relatively defect free materials?
Roger Newman answered: My main point is that the standard of proof is very
high if one wants to attribute an improvement in pitting resistance (by alloying) to
an increased resistance to passive lm breakdown, other parameters such as
surface nish and inclusion population, nature and size being kept constant.
I don't know of any instance where that standard has been met, whereas for
“dirty” alloys, the Galvele type of approach meets it routinely. You are asking
whether for clean and/or specially formulated alloys the lm might start to
dominate and could be engineered for resistance to breakdown. One response to
that would be to note that if pitting is not eliminated completely, the pits that do
initiate will grow rapidly – this is like “dangerous” inhibition. That is why I would
prefer to see passive lms engineered to reduce the rate of the cathodic reaction –
this is like “safe” inhibition. Assuming that in some applications the oxide could
be appropriately engineered on a very clean and smooth substrate (and one would
have to consider joining of the material, oxidation, surface damage, surface
cleanliness, plastic deformation, etc.), I have no doubt that the nucleation
frequency of pitting could be kept very low. We saw that ourselves in stainless
steel thin lms prepared by ion-assisted ion-beam sputtering (see Ryan citation in
another answer). Of course that is a kind of coating, so one might more logically
see improvement via the oxide route as something to pursue using advanced
vacuum coating technology applicable to large components. As you know, such
considerations apply differently to iron-base and aluminum-base systems, so
there is no commonality across the range of our usual materials that suffer from
pitting. It has been shown that thin-lm aluminum contaminated by oxygen has
Dirk Engelberg opened a general discussion of the paper by Steven Street: For
your time-lapse imaging experiment (Fig. 4) do you expect a change in relative
humidity over time? Do you take the sample out of the humidity cabinet, and is
there then enough time to equilibrate the environment before you take the next
images?
Published on 01 July 2015. Downloaded on 01/07/2015 16:00:38.
Geraint Williams asked: What role does differential aeration play in the pitting
corrosion of the 304 observed under a magnesium chloride containing water
droplet? For example, if the surface were to be fully immersed in an aqueous
solution of equivalent concentration, would spontaneous pitting still be observed.
Is there a lower threshold magnesium chloride concentration observed under a
droplet for spontaneous pitting in the absence of external polarisation, than when
the surface is fully immersed?
Is there any evidence that local alkalinity at the circumference of the droplet
causes precipitation of magnesium hydroxide in the region? Maybe this would
explain the ring of apparent metal excavation at the periphery of the droplet as
shown in Fig. 4.
Finally, for droplets of different sizes, but xed concentration, what parameter
controls the rate of anodic dissolution? If the pit growth rate is a function of the
area of solution in contact with the metal then this would indicate surface/
cathodic control rather than there being any ohmic constraints.
Steven Street responded: As noted in another response, the fact that pits do not
grow only in the centre of the droplets suggests that the entire surface is
vulnerable to pitting initiation in thin droplets. We have started exploratory
comparisons with full immersion experiments and have found pitting in
concentrations similar to those maintained at the RH used in our atmospheric
experiments. This work is in ongoing. Local alkalinity is thought to cause the
precipitation of Mg(OH)2 and possibly MgCO3, pinning the droplet perimeter.1
Figure 4 shows the perimeter of the pit, not the droplet, so we would expect the
solution still be to quite acidic here. For xed CDD and RH (concentration),
increasing the droplet diameter tends to give a larger shallow dish diameter, as
noted in another answer, so it is likely that this effect is due to the increasing area
available for the cathode reaction. Similar increases with droplet size were
1 T. Tsuru, K. Tamiya, and A. Nishikata, Formation and growth of micro-droplets during the
initial stage of atmospheric corrosion, Electrochim. Acta, 2004, 49, 2709–2715.
2 N. Mi, S. M. Ghahari, T. Rayment and A. J. Davenport, Use of inkjet printing to deposit
magnesium chloride salt patterns for investigation of atmospheric corrosion of 304
stainless steel, Corros. Sci. 2011, 53, 3114–3121.
Gerald Frankel asked: In our work on corrosion of SS under droplets at low RH,
the shallow pits smoothly transitioned into attack at smaller areas we called ears,
which we attributed to cathodic limitations.1 Why is that not a good explanation
for your observations of pit morphology change?
Published on 01 July 2015. Downloaded on 01/07/2015 16:00:38.
1 Bastian Maier and G. S. Frankel, Pitting corrosion of bare stainless steel 304 under chloride
solution droplets, J. Electrochem. Soc., 2010, 157, C302–C312.
Steven Street answered: The spiral shaped pits in this work are caused by the
“earing” mechanism you described, and this should indeed have been claried.
We agree that the role solution thickness and cathodic current limitation is
crucial to the development of these pits, which is supported by our observations
of the effects CDD and conductivity (RH). We extended this approach to include
the role of mass transport limitation, as we believe this is necessary in explaining
the formation of satellite pits. It is interesting that you did not observe satellite
pits, possibly due to a combination of slightly higher CDD and ner grinding of
the metal surface, which would have removed occluded initiation sites that in our
work led to satellites.
Dirk Engelberg asked: Is there interaction between satellite pits and the
shallow dish/pit region, since the shallow region seems to have grown into
satellite pits (e.g. Fig. 3 and Fig. 4)? (how does this affect cathodic vs. anodic
regions?)
David Williams said: I wonder about the temporal sequence of events in your
experiments. Could it be that the shallow dishes form as a consequence of the
activation of the surface near a pit by corrosion products (specically suldes or
thiosulfates) that are diffusing out of an active pit? Such effects have been shown
before.1
Cem Örnek commented: In 304L austenitic stainless steel the delta ferrite
content is about 6–14%, depending on the microstructure processing orientation
(TD, RD, ND). The ferrite has signicantly different chemical composition in
contrast to the austenite matrix which questions the role in localised corrosion.
Furthermore, there are MnS and other non-metallic inclusions oen present in
the microstructure. How are these related to the corrosion morphology during
atmospheric corrosion? Are there any preferential initiation sites observed? Can
we say from selective corrosion? Can we speak about pitting corrosion if we see
odd shaped corrosion morphologies?
Steven Street responded: Our steel contains ~4% delta ferrite, most of which is
below the plate surface. We have demonstrated that it inuences pit propaga-
tion.1 Our steel also contains sulphide and oxide inclusions which are known to
be pit initiation sites. However we have not carried out a systematic study on the
inuence of different initiation sites.
Cem Örnek asked: Evans introduced the atmospheric corrosion model under
thin-lm electrolytes formed by a hemispherical water droplet. Thus the anode is
primarily located in the centre of the droplet due to rapid oxygen depletion and
impeded oxygen supply and the periphery of the droplet forms the cathode due to
the excessive and quick access of oxygen. Preferential corrosion would be
expected to occur at or close to the centre due to the development of a large net
anodic potential. However, the results shown (and the poster from Angus Cook,
University of Birmingham) clearly contradict the model of Evans. Research on the
atmospheric corrosion behaviour on the same material and on 2205 duplex
stainless steel by my humble self can conrm your observed results. However,
there is no clear explanation given as to why the corrosion initiation occurs mostly
between the centre and the peripheral area of a hemispherical chloride-con-
taining droplet. Furthermore, the initiation of corrosion is concentrated on a
conned region, and, usually, only one or a few corrosion pits are observed which
is signicantly different to corrosion initiation sites under immersed conditions.
Roger Newman replied: I don't believe the situation is as tidy as the Evans
experiment. There are two major differences. First, the stainless steel is spontane-
ously passive, while the iron is spontaneously active. In the Evans experiment it is
the alkalization at the periphery of the drop that passivates the iron and creates the
active–passive couple. In the stainless steel experiment there is pit initiation and
propagation, which is very different. Second, oxygen reduction on stainless steel is
normally activation controlled, except where the stainless steel is in the active state,
while in the Evans experiment oxygen reduction is diffusion-limited. As I indicated
in another answer, oxygen reduction on stainless steel can also show apparent
“limiting current densities” that are too low to be diffusion-limited in the electrolyte.
Published on 01 July 2015. Downloaded on 01/07/2015 16:00:38.
Steven Street answered: Thank you for your very interesting comments. A
further brief discussion on Evans’ droplet model has been addressed in answer to
another question.
Dirk Engelberg asked: Can you suggest a viable corrosion mitigation strategy to
prevent localised corrosion in this system (since you show pitting corrosion occurs
with exposure to deliquescence relative humidity, and increasing the relative
humidity (e.g. 56%) also results in pitting corrosion). Do you accept a chloride
deposition density of 10 mm cm2 as a working threshold or have you observed
corrosion occurring below this threshold? Can you speculate whether a change of
shape during pit growth can be used to affect (and control) pit growth kinetics?
Steven Street responded: The obvious solution is to try to keep chloride concen-
trations as low as possible and to limit the thickness and continuity of the electrolyte
layer. An alternative method would be maintaining a very high RH in storage, to
ensure any salts that do deliquesce remain dilute, but these would necessarily pass
through an initial highly concentrated stage which may cause corrosion, if only
supercially. As initiation sites are seen to not reinitiate pitting once passivated,
natural wet–dry uctuations could be used to exhaust these sites on the surface of the
alloy.1 An improved alloy, such as 316L, would certainly improve performance. 10 mm
cm2 CDD is signicantly lower than the values used in this paper. A recent study2
suggests that there is merit to the use of a working threshold.
David Williams asked: I wonder if there are any convection effects associated
with surface tension gradients developed as droplets evaporate? Thus, are there
any effects that are different at the centre compared with towards the edge of the
droplet?
Steven Street replied: Convection effects are something we did not consider
Published on 01 July 2015. Downloaded on 01/07/2015 16:00:38.
and perhaps should have, as the droplet loses up to 95% of its initial volume as it
equilibrates with RH within the rst hour. We show in Figure 8 that the diameter
of the shallow dishes are greater in the centre of the droplet. We attribute this to
an increased IR drop and mass transport limitations at the edges. We do not
observe any other effect that might be attributed to convection.
Siaw Foon Lee communicated: Table 2 – the humidity of 33% has a droplet
area on deposition of 0.048 cm2 and has an average droplet CDD of 80 mg cm2.
The humidity of 43% has a droplet area on deposition of 0.048 cm2 that is same as
that of the humidity of 33%, however, it has an average droplet CDD of 790 mg
cm2. I am curious to know how the author works out the calculation since this
topic is new to me.
Stephane Heurtault replied: The ow rate of the solution is one of the main
parameters for the supply of chloride in the environment and it inuences the pit
growth. The supply of chloride (in mol h1) can be dened as the product of
chloride concentration (in mol L1) and ow rate (in L h1). However, we have
shown that we can control the amount of chloride independently by changing the
chloride concentration or by modifying the ow rate (see for instance the results
presented in Fig. 5). With a chloride ow rate of 5.4 mL h1 (as used in most of our
experiments presented in this paper) and taking into account the evolution of the
pit volume plotted in Fig. 10, only 3 min are required to renew entirely the pit
solution aer 1 h of propagation, and it will take 30 min aer 10 h of propagation.
Thus, the convection induced by the micro-capillary may be an important factor
that controls the pit propagation. To avoid such a behavior, it is possible to
generate and to propagate the pit in two successive steps: rst, the pit is generated
beneath the micro-capillary and then, it is sustained by controlling the bulk
concentration in chloride as was shown possible in the last part of the article.
We believe that the pit cover shown in Fig. 6(a) was not damaged during the
experiment but aerwards during transfer to the SEM. Indeed, in another
Roger Newman asked: Why do you use such a strong sulfate solution? It seems
that complications from the mixing of chloride and sulfate could be minimized if
you just reduced the sulfate concentration by a factor of 10. Some of your
experiments where you vary the chloride might be affected by varying degrees of
inhibition by sulfate (since inhibition is related to the chloride/sulfate ratio).
pit growth. In our previous work,1 differences between sulfate and perchlorate
solution were investigated. Moreover, some experiments were performed using 6
M SO42 and 1.2 M Cl instead of 0.5 M SO42 and 1.2 M Cl. With this increase in
sulfate concentration (corresponding to a decrease of the chloride to sulfate
ratio), the inhibitor role of sulfate was shown. We agree with this comment that
the chloride to sulfate ratio is an important point instead of only indicating the
chloride concentration.
Gerald Frankel stated: You show that the depth varies with t0.5, which would
also be found for ohmic control. How can you rule out ohmic effects?
Stephane Heurtault answered: Indeed it should have been mentioned that the
variation of pit depth with t0.5 also accounts for ohmic control as shown previ-
ously.1 The same dependence with time of pit depth for both ohmic control and
diffusion complicates the data interpretation. However for a pit under diffusion
control, potential variation has no effect on the pit growth rate, whereas a pit
growing under ohmic control depends on potential according Ohm’s law. Addi-
tional experiments at 20 C injecting 3 M NaCl + 0.5 M H2SO4 with the glass micro-
capillary were performed by changing the electrode potential during the pit
propagation (data not shown in this paper). The single pit was initially formed at
0 V/MSE. Aer 1 h of propagation, the potential was changed to 0.4 V/MSE. Then
the pit depth with this modication of potential was measured and the current
density at the pit bottom aer changing the potential was calculated using
equation (5). It was shown that the modication of potential does not affect the
calculated pit bottom current density, indicating a diffusion-control process.
pit bottom. Due to this salt lm, an additional potential drop at the pit bottom
exists, which implies a lower current density beneath the salt lm than at the pit
edges. These results are fully consistent with our experiments (disc-shaped pit
and salt lm at the bottom of the pit). The Laycock model also accounts for the
presence of a pit cover, which is composed of a central hole surrounded by many
small perforations forming the lacy-like structure. According to the model, the
cover entirely dissolves for long-term pit propagation. In our experiments, using
the adequate chloride concentration injected by the micro-capillary, we can get
single pits which propagate without any pit cover (this is possible according to the
Laycock model) or we can also observe the propagation of a single pit with cover.
However, the pit cover obtained in our experiments has only a central hole at the
location corresponding to the chloride solution injection by the micro-capillary,
and only one or two perforations can be observed in the periphery. This specic
behavior may be explained by the presence of the micro-capillary above the pit
maintaining the concentration of chloride ions at a high level. We didn’t see any
current uctuations resulting from the perforations of the pit cover. This was
conrmed by SEM observations. However for both covered and uncovered pits, we
noticed small current oscillations which can probably be ascribed to the growth
and the breakdown of a thin oxide layer under the salt lm.1,2
Once the pit propagates, we have never observed repassivation aer the
collapse of the pit cover. However, it should be mentioned that an increase of the
electrode potential during the pit propagation was accompanied by the collapse of
the pit cover. In that case, the current increased sharply during the collapse, but
the propagation was maintained.
U(233)) to provide a known dose rate have been used.1 This method provides an in
situ a-radiation eld.
(c) An alternative method for a-radiolysis is to use a thin layer electrochemical
cell with the electrode (UO2) brought close to a Au-coated a-source (G 25mm) with
a known dose rate.2 This method provides an ex-situ source.
(d) Intense beams (e.g., X-rays (Ross Springell et al.3)) can be used, but usually
produce extremely high dose rates well beyond those expected inside a failed
container.
(e) The inuence of g-radiolysis (and also a-radiolysis) can be simulated by
adding the anticipated concentration of hydrogen peroxide chemically. Evidence
in support of this approach is well documented.4
Ross Springell replied: Simulating the effects of corrosion of UO2 under real-
istic conditions initially requires one to dene what these conditions might be,
and this can depend on what scenario you choose. The most frequently used
method is the addition of small quantities of H2O2 to the aqueous environment.
This is based on the assumption that any container of spent nuclear fuel will likely
fail aer the rst few hundred years when the gamma elds from the ssion
products have decayed and the dominant radiation elds are driven by alpha
decay; in this case the dominant radiolytic product will be H2O2. However, the
most likely cause of any containment failure is due to potential defects in the
vessel weld. This will be present from day 1 and it is possible that water may
ingress way before all of the ssion products have decayed. In this instance,
strong gamma elds will also be present and this will result in both H2O2 and
other more oxidising free radicals at the fuel surface. A lot of this discussion really
depends on the precise packaging methods and eventual repository conditions.
One method to better simulate the radiation elds would be to study changes in
the UO2/water interfaces exposed to an alpha source and a gamma source, but
these would have to be ex-situ studies with characterisation of the structure with
David Shoesmith answered: The primary basis of a model like this one (fuel
corrosion model) is a mechanistic understanding of the key reactions required to
describe the corrosion process. Our guiding principle is that the model should
contain the minimum number of reactions necessary to reasonably represent the
Published on 01 July 2015. Downloaded on 01/07/2015 16:00:38.
overall process. To achieve this the model should contain the minimum number
of adjustable or disposable parameters. Ideally, the model should then yield
output values which can be compared to trusted experimental data. Unfortu-
nately, required data is commonly not available, unreliable, or determined for
inappropriate conditions. Disagreement between calculated (model) and
measured values can, however, provide a template for required measurements or
theoretical calculations.
Unfortunately, models and calculations can be developed and performed on a
time scale which cannot be matched by experimental determinations. This
commonly leads to the “over-extension” of models beyond validation, either by
comparison to measurements or by theoretical calculations. This is generally the
reason for the use of adjustable parameters. The more such parameters are
included the more over-extended and unreliable the model becomes. Such over-
extensions should be minimized by the iterative process of the model (oen
sensitivity) calculations, experimental determinations, model redevelopment.
Alison Davenport asked: Could you give further details on the uncertainty in
the input parameters to the model, and to the sensitivity of the model to different
parameters? Are there any input parameters for which improved measurements
would be helpful to reduce the uncertainty in the model?
1 L. Wu, Y. Beauregard, Z. Qin, S. Rohani and D. W. Shoesmith, Corros. Sci., 2012, 61, 83–91.
2 Linda Wu, Zack Qin and David W. Shoesmith, Corros. Sci., 2014, 84, 85–95.
3 Linda Wu, Nazhen Liu, Zack Qin and David W. Shoesmith, J. Electrochem. Soc., 2014, 161,
E3259–E3266.
Christopher Taylor asked: Early in the presentation you mention the hypoth-
esis or expectation that the epsilon particles, being noble, could drive anodic
processes by acting as a cathodic electron sink by catalyzing HER. However, it
appears that instead, your expectation is that they will have an opposite effect, by
Published on 01 July 2015. Downloaded on 01/07/2015 16:00:38.
quenching H2O2. Is this the correct interpretation of the role of epsilon particles,
or does the competition between these two effects still need to be elucidated?
Ross Springell said: You mention in your article that UO2 has the potential to
have a catalytic effect that may inuence its corrosion. Could you expand on this
point and describe how important the stoichiometry is in this aspect?
Dirk Engelberg said: With the presence of hydrogen peroxide in the aqueous
phase, and/or catalytic reactions expected to occur at the uranium surface/parti-
cles (H2/H+) wouldn't you expect the formation of uranium hydride?
David Shoesmith answered: The fuel form is uranium dioxide not uranium
metal. Hydride formation would only be expected if uranium in the zero oxidation
state was present. While H atoms will form on the noble metal particles (an alloy
of Ru, Pd and Mo), and some minor absorption into the particle may be possible,
this will not lead to reduction of the U(IV) oxide to U metal.
Christopher Taylor remarked: For the water gas shi and reverse water gas
shi research efforts, bifunctional catalysts that contain nanoparticles of precious
or noble metals dispersed over a redox-active oxide, like ceria, are used to tune the
reaction chemistry for CO2 + H2 # CO + H2O and potentially form hydrocarbon
products. In your case, you have a similar set-up between these epsilon particles,
which are noble and catalytic, and the redox active UO2. By analogy, then we
might expect the interfaces between the particles and the urania to be the active
regions whereby CO2 is able to engage in chemistry with the material. Are there
other possible insights that could be gained by this analogy?
David Shoesmith responded: The analogy is a good one. In the reaction you
describe on ceria you note that the whole oxide matrix is involved not just the
dispersed noble metal nanoparticles, but you suspect the coupling within the
uranium dioxide will be limited to regions close to the noble metal particles. I
agree. The area of the reactive uranium dioxide inuenced by the noble metal
anode/cathode will be determined by the electrical conductivity of the rare earth
doped oxide lattice and the ionic conductivity of the electrolyte it is exposed to;
i.e., this is a normal microgalvanic (or internal) coupling process.
I think the analogy can be extended in terms of the behaviour of the uranium
and cerium oxides. The hydrogen peroxide we are interested in will react directly
on the uranium dioxide surface besides via the galvanic coupling process, making
the catalytic properties of the oxide also important. Both cerium oxide and
Clara Wren opened a general discussion of the paper by Ross Springell: The
concentration of the radiolytically-produced H2O2 that you quoted is very high at ~
104 M. What was the energy absorbed by water? Can X-ray induced radiolysis
create that high an H2O2 concentration? The photoelectrons generated in the UO2
matrix should lose most of their energy to overcome the binding energy and work
function, and, when they are ejected from the UO2 matrix, will have a relatively
low energy, on the order of ~ 100 eV. This is not sufficient to penetrate the
Published on 01 July 2015. Downloaded on 01/07/2015 16:00:38.
adsorbed water layer but is sufficient to ionize water molecules close to the UO2
surface, affecting the water chemistry close to the surface. If this is the case, the
X-ray induced corrosion may be different from that driven by higher energy
radiation such as gamma-rays. Can you comment on the potential differences
arising in X-ray and g-ray induced corrosion?
Ross Springell replied: The X-ray energy is 17 keV and for the average 0.5 mm
coverage of water that we used in this experiment, the transmission was 59%,
which means that 41% of the total photon ux (approx. 1012 photons s1) was
absorbed in the beam track through the water. By using the linear energy transfer
(LET) of X-rays at this energy it is possible to calculate the number of H2O2
molecules produced per 100 eV deposited, per photon. The high photon ux
results in the increase of H2O2 concentration up to its equilibrium concentration
within a fraction of a second. Both the X-rays and gamma rays have a similar LET
and produce similar proportions of radiolytic product, but it is certainly the
situation that the absorption in the water is much stronger for the X-ray case. In
fact, it is this property that we rely on to enable us to measure the corrosion in situ
during the time frame of a synchrotron experiment; in real spent fuel stores the
rates of corrosion are considerably slower.
Geoffrey Thornton stated: We have recorded SXRD data (CTR's) from a water/
TiO2(110) interface using ID32 at the ESRF. The photon ux was also 17 keV, with
a photon ux of 1012 ph s1. We see an interface ordered on the atomic scale. In
other words, there is no evidence of beam damage. A key difference between your
sample and the TiO2(110) substrate is that the former employs an insulating
substrate, whereas the latter was a self-doped n-type semiconductor. Hence, it
may be a charging effect that gives rise to the beam damage you observe.
planning an in situ high speed AFM study of the corrosion of a number of UO2
crystal grains on a longer time-scale, using hydrogen peroxide as the oxidant.
David Shoesmith asked: Have you tried to characterize the UO2 layer before
exposure and the oxidized surface layer aer exposure using techniques such as
Raman spectroscopy and X-ray photoelectron spectroscopy?
Philippe Marcus asked: Can we really implant in the human body a stent that
would produce a lot of hydrogen? What is the conclusion in terms of the effect of
iron impurity in magnesium on hydrogen evolution (one of the hypotheses
included in the Introductory Lecture presented by Gerald Frankel1 to explain the
“anodic” hydrogen) and what are the implications for the application of Mg alloys
for stents (requested purity of Mg)?
Geraint Williams responded: This point is well made. The majority of research
devoted to bioresorbable Mg implants largely ignores the fact there may be
problems which arise in vivo due to high local concentrations of hydrogen caused
by corroding magnesium. Most attention has been concentrated on the
biocompatibility of the cations which are released upon corrosion. Given the well
documented inuence of iron impurity level on corrosion rate, closely controlling
the trace Fe content of Mg may provide a means of “tuning” the corrosion rate to
one which lies in an acceptable time window for optimum in vivo performance.
However, in reality the situation will be much more complicated since it is likely
that nominally pure Mg, with a certain “tuned” Fe trace content, may not possess
the requisite mechanical strength (e.g. for stent applications), and would need to
be alloyed with other elements. There may, in turn, be some interplay of these
additional alloying elements with Fe, which could either mitigate (e.g. Zr, Al–Mn)
or enhance (e.g. Si) the inuence of Fe on corrosion rate. As a result, there are
probably more straightforward methods of limiting corrosion rates, such as the
use of an appropriate biocompatible surface treatment or coating.
Stuart Lyon remarked: Should we be much more concerned with the inuence
of “low” or trace concentrations of elements in alloys and their inuence on the
surface structure/chemistry and corrosion?
John Scully replied: As mentioned in the closing lecture,1 I believe that trace
and minor alloying element effects are both one of the great triumphs of the
corrosion metallurgy eld and one of its least well understood aspects.
Consider the effects of Mo and nitrogen on stainless steels as well as Mo and W
on Ni–Cr alloys. Consider minor and trace elements effects on corrosion in copper
base alloys. Consider the effects of Cu and Ni or others in weathering steels. In the
case of amorphous alloys, minor alloying additions are oen made to improve
quench rate and glass formation but effects on corrosion are not well understood.
The need for understanding intensies further in high entropy alloys.
These effects are not well understood yet provide some of the most productive
area of technological accomplishments in corrosion metallurgy. Therefore, I see
some of these topics as very worthy for investigation as corrosion science looks
towards the future.
Edisonian.
Regarding ICCME (Computational Corrosion Materials Eng), the success
stories are limited and in isolation but include tunable Al-TM-RE glass and
nanocrystalline coatings that our group invented,1–4 ultra high strength stainless
steels that do not require heavy metal coatings (Questek), and iron based BMG
which, maybe by accident, combine all the tools in the MSE tool box to make very
corrosion resistant materials (signicant Cr and Mo in solid solution, benecial
minor alloying elements such as C, B, Y, lack of inclusions or large defects etc.).
Another example might be grain boundary engineering of polycrystalline alloys
for SCC resistance. These examples stand in isolation and do not yet signal a
broad movement.
1 Asmae Bouzoubaa, Boubakar Diawara, Vincent Maurice, Christian Minot and Philippe
Marcus, Corros. Sci., 2009, 51, 2174–2182.
2 Anton Kokalj, Faraday Discuss., 2015, DOI:10.1039/c4fd00257a.
3 Bai-Lin Lü, Guo-Qing Chen, Wen-Long Zhou, Hui Su and Rui Liu, J. Nucl. Mater., 2011,
286–291.
4 Toni Massoud, Vincent Maurice, Lorena H. Klein and Philippe Marcus, J. Electrochem. Soc.,
2012, 160, C232–238.
5 Kiran B. Deshpande, Electrochim. Acta, 2011, 56, 1737–1745.
6 Nicolas Murera, Roland Oltra, Bruno Vuillemin and Olivier Néel, Corros. Sci., 52, 130–139.
7 Dao Trinh, Philippe Dauphin Ducharme, Ushula Mengesha Tefashe, Joseph R. Kish and
Janine Mauzeroll, Anal. Chem., 2012, 84, 9899–9906.
8 Daniel Höche, J. Electrochem. Soc., 2015, 162, C1–C11.
Published on 01 July 2015. Downloaded on 01/07/2015 16:00:38.
Alison Davenport asked: The growth velocity of the corrosion laments can be
converted into a current density. Could you comment on the range of current
density values found, and in particular if there is a minimum current density
below which lament growth cannot take place?
Gerald Frankel asked: It seems that you should be able to understand the
phenomenon by assessing the cathodic activation aer corrosion using polari-
zation curves. This might vary with the chloride concentration similar to the effect
of the prior applied anodic current density. In that case, the corrosion
morphology might be controlled by the amount of available cathodic current and
galvanic coupling rather than by the ohmic resistance.
present work disseminated in this Faraday discussion. The random walk FFC
patterns observed upon immersion in 0.01 M electrolyte (see Fig. 1c), for the
rather impure form of Mg studied in our work, suggests that microstructural
inuence on lament directionality is minimal in the absence of dissolved
chromate inhibitor.
1 S. Thomas and H. Bohni, Analytical Methods in Corrosion Science and Engineering. CRC
Press, Boca Raton, FL, USA, 2005, pp. 649–696.
2 Williams et al., Anal. Chem., 2009, 81, 2486–2495.
1 K. A. Lill, A. W. Hassel, G. Frommeyer and M. Stratmann, Electrochim. Acta, 2005, 51, 978–
983
2 M. M. Mennucci, M. Sanchez-Moreno, I. V. Aoki, M. C. Bernard, H. G. Melo, S. Joiret, and
V. Vivier, J. Solid State Electrochem., 2012, 16, 109–116.
3 M. M. Lohrengel, A. Moehring and M. Pilaski, Fresenius’ J. Anal. Chem., 2000, 367, 334–339.
4 M. Pilaski and M. M. Lohrengel, Electrochim. Acta, 2003, 48, 1309–1313.
5 F. Zou and D. Thierry, Electrochim. Acta, 1997, 42, 3293–3301.
6 D. Snihirova, L. Liphardt, G. Grundmeier and F. Montemor, J. Solid State Electrochem.,
2013, 17, 2183–2192.
7 A. M. Mierisch and S. R. Taylor, J. Electrochem. Soc., 2003, 150, B303–B308.
8 M. W. Wittmann, R. B.Leggat, S. R. Taylor, J. Electrochem. Soc., 1999, 146, 4071–4075.
9 W. Liu, F. Cao, Y. Xia, L. Chang and J. Zhang, Electrochim. Acta, 2014, 132, 377–388.
10 U. M.Tefashe, M. E. Snowden, P. Dauphin-Ducharme, M. Danaie, G. A. Botton and
J. Mauzeroll, J. Electroanal. Chem., 2014, 720–721, 121–127.
11 P Dauphin Ducharme, R. M. Asmussen, D. W. Shoesmith and J. Mauzeroll, In situ Mg2+
release monitored during magnesium alloy corrosion, J. Electroanal. Chem., 2015, 736,
61–68.
12 Y. Lee, S. Amemiya and A. J. Bard, Scanning electrochemical microscopy. 41. Theory and
characterization of ring electrodes, Anal. Chem., 2001, 73, 2261–2267.
Rob Lindsay said: From reading your paper, I believe you have assumed a
homogenous substrate for quantication of the XPS data. Clearly, if this is true,
this substrate model does not represent reality particularly well, in my opinion, as
there will be discrete layers of material, with the metallic substrates lower most.
How do think your approach impacts upon your interpretation of the XPS data
(e.g. data shown in Fig. 8)?
Sannakaisa Virtanen replied: The XPS data presented simply shows the at%
determined, and is not further interpreted for instance in view of a homogeneous
or partial coverage or in view of the different layers from the metallic substrate, to
Published on 01 July 2015. Downloaded on 01/07/2015 16:00:38.
passive lm, to the protein adsorption layer. We do not believe that we have a well-
ordered and complete monolayer of albumin in this case. The XPS data should be
considered only in a qualitative manner, i.e. in comparison of the different
systems (substrate material, time, electrolyte). In order to perform a full analysis,
more experimental data would be required, including information on coverage
and thickness of the different layers.
Sannakaisa Virtanen replied: Thank you very much for the important
comment and helpful literature. Yes, indeed not only the type of surface charge
but also the charge density should play a role. Until now, we did not carry out any
detailed studies on this, and the surface charge considerations in our paper are
simply based on IEP values found in the literature.
But indeed we would like to know more about the type of the protein
adsorption layers, so in addition to addressing the question on the amount of
protein adsorbed on different surfaces, we would like to study the orientation of
the proteins on the surface (with additional experiments and a more detailed data
analysis) and also their functionality (for instance, are they in their native
conformation or possibly denaturated, depending on the surface they are
adsorbing on).
In the case of Mg some of the studies we would like to do are challenged by the
high reactivity of the Mg surfaces in aqueous solutions, so we oen have the
problem of not only adsorption of proteins taking place but at the same time
substantial corrosion of the substrate (this is also briey mentioned in the paper).
Sannakaisa Virtanen communicated in reply: Thank you for the very inter-
esting comments and questions and also for sharing some unpublished data. I
completely agree that our approach is a rough and simple one, as certainly the IEP
values should be measured for the sample surfaces under the conditions of the
study, to have a better and more realistic view. Also, I agree that surface charge is
not the only factor inuencing protein adsorption, for instance surface wettability
is also of importance. If other driving forces are stronger than electrostatic
interactions, then adsorption could take place as you mention also for electro-
statically unfavorable conditions. In our case, as both Fe and Mg surface are
hydrated and show a similar hydrophilic behaviour, according to the literature
electrostatic interactions can be more signicant.
As regards the nature of the Fe surface: The passive lm formed in NaOH
should be an oxide with a mixed valency between Fe3O4 and Fe2O3 (but not FeO).
We refer to the values for FeO/Fe(OH)2 in Table 2, as we assume a hydroxide
deposit layer on top of the Fe-oxide layer. Again, a determination of IEP values for
the given case should be carried out.
The discussion in the manuscript is based on the fact that our hypothesis on
the relevance of the surface charge argument seems to hold for two different
proteins. But it still remains a hypothesis.
Clara Wren opened a general discussion of the paper by James Sullivan: Your
corrosion rate decreases with the aluminium content in the alloy. Have you tried
to normalise the corrosion rate with respect to the same zinc–magnesium or
magnesium content? Is the change in corrosion rate simply a function of the
relative amount of magnesium available, or is there any synergistic effect of the
composition-induced alloy-microstructure on the corrosion rate?
Stuart Lyon asked: What is the specic composition of the various components
of the eutectic in these complex Zn–Al–Mg alloys? How does the area/volume
fraction of eutectic inuence the overall corrosion performance?
Published on 01 July 2015. Downloaded on 01/07/2015 16:00:38.
Gerald Frankel said: Why doesn't the attack in the eutectic region go deep into
the structure and why does it initiate at only a few places and then spread across
the surface?
James Sullivan responded: The attack does also penetrate vertically into the
sample and we have seen this from examination of cut samples post corrosion
although it would seem that the rate of lateral spreading is greater. The reason for
the corrosion spreading through the eutectic laterally at the surface may be due to
the surface being more eutectic rich as the solidication sequence starts with
nucleation of Zn rich dendrites at the steel substrate. Therefore the volume
fraction of primary Zn is greater at the coating/steel interface with a subsequent
reduction of primary Zn and increase in eutectic as the coating is traversed to the
surface. As the MgZn2 in the eutectic is the preferential phase for initial corrosion
attack then lateral attack through this eutectic rich surface may dominate.
Arjan Mol asked: First, I would like to thank James Sullivan for presenting this
very interesting work and paper during this Faraday Discussions session, I greatly
enjoyed it. My question is the following: can you please show, explain and/or
comment on the positions and evolution of (local) cathodic and anodic areas in
the SVET measurements with time and their correspondence to those in the time-
lapse optical microscopy videos considering a.o. the different lateral resolutions
of the techniques?
Steven Street asked: The three different ZMA alloys had quite different
dendrite sizes, indicating different thermal gradients when solidifying. As
corrosion seemed to propagate in the interdendritic regions, is there any value in
trying to rene the size of these dendrites? Would this improve the corrosion
behaviour of these alloys?
1 J. Elvins, J. A. Spittle and D. A. Worsley, Corros. Eng., Sci. Technol., 2003, 197–204.
Stuart Lyon asked: Corrosion resistant metallic coatings on steel are much
thinner than they used to be. How does this inuence the corrosion process/
resistance at the cut edge?
James Sullivan answered: These ZMA coatings are certainly thinner than more
traditional HDG type coatings but in industrial salt spray tests they certainly
outperform the HDG materials during surface testing. We also see this
improvement in corrosion performance when examining surface corrosion. At the
cut edge we have comparable performance to that of HDG for the thinner ZMA
coating in SVET and external run-off testing that is encouraging considering the
change in anode/cathode ratio. However, the performance at the cut edge is not
nearly as improved as the surface performance in comparison to HDG, that may
reect this change in anode/cathode ratio.
Clara Wren remarked: What is the geometry of the viewing cell, the diameter
and the depth of the solution, with respect to that of the rest of the corrosion cell?
I noticed in your video images that the corrosion front always starts close to the
boundary between the viewing area and the rest of the corrosion cell. Could your
observation indicate that the Mg corrosion is solution transport limited?
James Sullivan replied: I too feel that this is potentially showing some inter-
esting information about the nature of anodic and cathodic sites, the ionic
diffusion between the two electrodes and pH gradients established between them.
The pH effects have been demonstrated in some previous work1 using a
phenolphthalein indicator and is visible due to the dissolution of Mg corrosion
products as the front of anodic activity moves, presumably as the pH drops local
to the anodic site. I think it may also be interesting to see the effect of electrolyte
concentration on the formation of the corrosion product rings with regards to the
distance from the anode to assess the effect of the electrolyte's throwing power on
anode-cathode spacing.
1 James Sullivan, Shahin Mehraban and Jon Elvins, Corros. Sci., 2011, 53(6), 2208–2215.
Su-Ting Cheng said: In Fig. 2, the red area increases largely from ZMA1 to
ZMA3. Since the bulk concentrations of Zn (green) and Mg (red) are kept the
same, a different colour distribution could indicate a surface segregation of Mg.
Have you ever measured the concentration of elements throughout the whole
depth of the specimens? We have studied the Mg–Zn system with DFT, and the
results tell us the surface of the MgZn2 intermetallic compound is more stable
with a higher surface Mg concentration, either in vacuum or in contact with
oxygen. It would be nice to see if this could be also observed in experiments.
James Sullivan responded: I may have misunderstood the question but the
quantity of Mg increases from 1–3 wt% from ZMA 1–3 and the Zn level drops as
the Mg and Al levels increase. This I believe accounts for the increase in red as the
volume fraction of eutectic increases in the microstructure from ZMA 1–3. At
present we have not measured the depth prole of the elements.
application of point defect models to give new insights into the defects formed on
zinc surface which could lead to corrosion. In your paper you have considered the
corrosion behaviour of commercially important zinc alloys. I would be interested
to know if you have considered applying such models to help in the under-
standing of your systems?
James Sullivan answered: This is not something we have considered to date but
it would be of interest to assess how the model could be applied to these alloys to
provide a prediction of corrosion behaviour with regards to preferential attack.